0.25
0.5
0.75
1.25
1.5
1.75
2
Semi-supervised multi-target prediction for analysis of screening data
Published on 2019-06-2879 Views
The predictive performance of traditional supervised methods heavily depends on the amount of labeled data. However, obtaining labels is a difficult process in many real-life tasks including compound
Related categories
Presentation
Semi-supervised multi-target prediction for analysis of screening data00:00
What is semi-supervised learning?00:34
Why semi-supervised learning?02:06
Outline - 103:08
The task of semi-supervised learning03:32
How unlabeled data can help?05:12
Why multi-target prediction?07:40
SSL for classification tasks10:44
SSL for regression tasks11:18
SSL for multi-label classification11:45
SSL for multi-target regression11:57
Existing SSL methods for SOP12:11
Limitations of the existing methods12:42
Outline - 213:43
Predictive clustering trees13:54
Supervised PCTs14:29
PCTs instantiations15:17
Semi-supervised PCTs16:06
Predictive clustering17:42
Ensembles of semi-supervised PCTs18:51
Outline - 319:20
Experimental evaluation19:24
Experimental setup20:26
Predictive performance (examples)21:07
Statistical analysis22:12
Influence of the w parameter22:53
Influence of the unlabeled data23:57
Interpretability and model sizes24:47
SSL-PCTs for primitive outputs25:47
Illustrative study on QSAR datasets26:26
Performance results26:57
Interpretability potential27:28
Obtained PCTs - 128:27
Obtained PCTs - 229:02
Obtained PCTs - 329:09
Obtained PCTs - 429:14
Obtained PCTs - 529:17
Outline - 429:22
Conclusions29:25